Question about RTE ring

Question about RTE ring

[arie abergel]

Hi,

As part of a project I have a question about the rte ring.
I’m using rte ring multi producer/single consumer.
The producers are several process.
If one producer is enqueuing an element and crashed (kill pid) in the middle of the
 enqueuing, can it compromise the ring ?

Thanks !

RE: Question about RTE ring

[Konstantin Ananyev]

 Hi,
> 
> As part of a project I have a question about the rte ring.
> I’m using rte ring multi producer/single consumer.
> The producers are several process.
> If one producer is enqueuing an element and crashed (kill pid) in the middle of the
>  enqueuing, can it compromise the ring ?

I suppose you are using rte_ring as IPC mechanism between multiple processes, correct?
In theory - yes, if your producer crashed during enqueue() to the ring, then yes, the ring might be affected.
If producer already moved prod.head  and crashed before updating prod.tail, then no other producers
will be able to enqueue() into the ring, till you'll do reset() for it.
I expect such situation really rare and hard to reproduce, but in theory it is possible.
Konstantin  

Re: Question about RTE ring

[arie abergel]

Thans for your response !

Yes, using rte_ring between multiple process.

So in this case you’re saying the behavior is undefined ?
In my case another process crashed after that.

> Le 11 avr. 2024 à 11:08, Konstantin Ananyev <konstantin.ananyev@huawei.com> a écrit :
> 
> 
> 
> Hi,
>> 
>> As part of a project I have a question about the rte ring.
>> I’m using rte ring multi producer/single consumer.
>> The producers are several process.
>> If one producer is enqueuing an element and crashed (kill pid) in the middle of the
>> enqueuing, can it compromise the ring ?
> 
> I suppose you are using rte_ring as IPC mechanism between multiple processes, correct?
> In theory - yes, if your producer crashed during enqueue() to the ring, then yes, the ring might be affected.
> If producer already moved prod.head  and crashed before updating prod.tail, then no other producers
> will be able to enqueue() into the ring, till you'll do reset() for it.
> I expect such situation really rare and hard to reproduce, but in theory it is possible.
> Konstantin  

RE: Question about RTE ring

[Konstantin Ananyev]

> 
> Thans for your response !
> 
> Yes, using rte_ring between multiple process.
> 
> So in this case you’re saying the behavior is undefined ?
> In my case another process crashed after that.

Without a proper debug session it is really hard to tell what is going on.
If the situation is reproducible, I'd suggest to run it with gdb and see.

> 
> > Le 11 avr. 2024 à 11:08, Konstantin Ananyev <konstantin.ananyev@huawei.com> a écrit :
> >
> > 
> >
> > Hi,
> >>
> >> As part of a project I have a question about the rte ring.
> >> I’m using rte ring multi producer/single consumer.
> >> The producers are several process.
> >> If one producer is enqueuing an element and crashed (kill pid) in the middle of the
> >> enqueuing, can it compromise the ring ?
> >
> > I suppose you are using rte_ring as IPC mechanism between multiple processes, correct?
> > In theory - yes, if your producer crashed during enqueue() to the ring, then yes, the ring might be affected.
> > If producer already moved prod.head  and crashed before updating prod.tail, then no other producers
> > will be able to enqueue() into the ring, till you'll do reset() for it.
> > I expect such situation really rare and hard to reproduce, but in theory it is possible.
> > Konstantin